The aim of this proposal is to investigate the possibility of detection and predicting relapse long before it becomes clinically apparent in children with B lineage acute lymphoblastic leukemia. The long-term objective is the improvement of therapeutic protocols for those children (20-30%) that respond poorly to current protocols. We will utilize a highly specific and sensitive method (CDR3-PCR) that we have developed for quantitation of normal and leukemic B cell clones (Yamada et al. PNAS, 1989; Yamada et al, J. Exp. Med, in press). This method utilizes diagnostic oligonucleotide probes that are homologous to the CDR3 region of the IgH locus in the leukemic cells to screen recombinant libraries containing the amplified CDR3 sequences from marrow B lymphocytes in order to determine the ratio of leukemic B cells two norm,al B cells. Preliminary studies (Yamada et al. NEJM, 1990) have indicated that during maintenance therapy residual leukemic cells are present at low levels, and that several months before a clinical relapse occurred in one patient a drastic increase in the number of malignant cells was detected in a remission marrow by this method. In particular, we now propose to determine 1) whether there is a correlation between the extent of cytoreduction by induction therapy and the chances of relapse 2) whether we can identify patients with a poor or failing response to maintenance therapy, and 3) whether the presence of residual leukemic cells in patients off therapy implies eventual relapse several years later.